Dmc Virtual Image Characterization: Experiences at Icc

It has been proven that DMC images (as images of other digital aerial cameras) are not free of systematic errors in the virtual image space. Not properly modelled estimated exterior orientation can absorb propagated errors in the bundle block adjustment (from different errors sources as image systematic errors, poor GPS/INS observations or insufficient ground control points set up) and could generate unwanted large systematic errors in the object space, especially in height. To keep error propagation under control two different approaches are considered in bundle adjustment: i) an appropriate set of self-calibration parameters and ii) calibration/characterisation grid, compensating image systematic errors in each virtual image. The calibration grid is derived from a calibration flight and should be valid for images acquired in other projects. This paper focuses on two topics: firstly a comparison between the performance of both approximations in aerotriangulation, and, secondly, the impact of these systematic errors on stereoplotting. The first analysis uses seven different data sets (including the calibration flight) and three bundle adjustment set ups: i) not using any model at all, ii) using a self-calibration parameter set, and iii) using the calibration gird without using any selfcalibration parameter set. Independent check points are used to assess the performance of both techniques in bundle adjustment. In the second analysis object coordinates for points of single models are calculated compensating and also not compensating for systematic errors their image coordinates. Later, these two sets of points are compared to the respective estimated object coordinates of the bundle adjustment. Main results of this work suggest that application of calibration grid (as it is derived in this analysis) is not able to isolate the systematic errors in virtual image from other errors sources in the calibration block. The comparison with selfcalibration indicates a small superiority of self-calibration, being less sensitive to GSD and seasonal time variations. In stereoplotting or DSM generation preliminary results show a small improvement using calibration grid compensation with respect to not compensating for systematic errors in image space. The effects of systematic errors in virtual image space can be much more significant in the computation of the exterior orientation in bundle block adjustment than in stereoplotting or DSM generation, because here the exterior orientation is given and the non modelled systematic errors are not propagated through the block.